Critical reevaluation of an efficient sampling design for assessing soil properties using bootstrap sampling and geostatistical analysis in Japanese large-scale paddy fields

Abstract

ABSTRACT Soil test is a key step toward providing recommendations for better crop management. Several soil samples have been traditionally assumed to be sufficient for soil tests to represent field-specific values in conventional Japanese small-scale paddy fields. However, rethinking soil sampling design is required, as many small-scale (<0.3 ha) paddy fields have been consolidated into large-scale (>1 ha) paddy fields to enhance the efficiency of crop production. The purpose of this study is to explore an efficient soil sampling design, including sample size for representing field-specific values and sampling distance for representing spatial variations, in central Japan using bootstrap sampling and geostatistical analysis. Fourteen soil properties were quantified from 553 samples, which was collected at a distance of 24.4 m on average in large-scale paddy fields with continuous rice cultivation and rotation of rice and upland crops (winter wheat and soybean). The results show that the conventional sampling size (n = 3 for each field) achieved mean estimation within 10% error with 95% confidence intervals only for pH and sand content in almost all fields; thus, an optimization of field-specific uniform liming rate is recommended for reducing cost. Geostatistical analysis shows that the recommended soil sampling distance should be 15–163 m, depending on specific soil properties. The results further show that it was difficult to obtain reliable estimates of exchangeable K and mineralizable N because of the high level of spatial uncertainty with high nugget variance. Thus, practitioners should note that the outcomes from soil tests inherently included fine-scale errors in available nutrient levels which may preclude rationale prescriptions. This study demonstrated that appropriate soil sampling design and the subsequent soil management can differ depending on specific soil properties in the actual farming scale of large-scale paddy fields.